Claude Max and Team get API credits that skip Claude Code

Anthropic's new monthly API credits for Max and Team subscribers work with any Claude model you call from your own code. They do not cover interactive Claude Code, the tool most of those subscribers open every day. The terms are in a help center article from Claude: Max 5x gets $100 a month, Max 20x gets $200, and a Team plan gets up to $500 in a shared balance.

At a glance

• Max and Team subscriptions now include a monthly balance on the Claude Platform, meant for building your own apps and agents with the Claude API, Claude Managed Agents and the Claude Agent SDK.
• Team seats earn $20 for Standard and $100 for Premium, pooled into one balance capped at $500, so three Standard plus two Premium seats bring in $260 a month.
• Pro subscribers are left out entirely, the balance lands in one Console organization only, and it cannot pay for interactive Claude Code or extra usage beyond plan limits.

If you missed the earlier rounds: in April, Anthropic temporarily banned third-party agent tools from subscriptions. On May 14, 2026 it announced that Agent SDK and claude -p usage would leave subscription pools on June 15, 2026 and move to a separate monthly credit at standard API rates, with no rollover. That plan gave Pro $20, Max 5x $100 and Max 20x $200, per user and not pooled. According to Digitalapplied, Anthropic paused the change on June 15 and said it was reworking the plan.

Max 5x gets $100 a month, Max 20x gets $200, and Pro is left out

The new credits replace the June design. Anthropic's help article says the Agent SDK credit announced in June isn't available, and the new API credits cover the Claude API, Claude Managed Agents and the Claude Agent SDK. Max 5x subscribers get $100 a month and Max 20x subscribers get $200, the same amounts the June plan promised.

Pro is where the plan changed most. In June it was set to get $20. The new terms answer the question directly: the credits are available on Max and Team plans only. The money goes to one Console organization, and you cannot split it across several.

The rollout takes a few days. The offer appears after you have been on an eligible plan for seven days, and you also need one of the required roles. Anyone past the seven days who still sees nothing is told to check their plan and role, then contact support.

A Team pool tops out at $500, however many seats you add

Team plans earn $20 per Standard seat and $100 per Premium seat. Discounted Team plans for nonprofits and scientists get the same amounts: $20 per Standard seat and $100 per Premium seat. All seats feed one monthly balance, capped at $500.

The pool is recalculated from the seats on the plan each time credits are deposited. In Anthropic's example, three Standard seats and two Premium seats yield $260 a month, and one more Premium seat raises the next cycle's credits to $360.

One owner claims the credit for the whole team by linking a Console organization. Everyone with an API key in that organization can spend from the balance. To limit spending by project or by teammate, owners set workspace spend limits in the Console.

Haiku 5.5 costs $0.10 per million input tokens on prompts under 100K

According to the ClaudeDevs account on Twitter, the credits work on any model, in your own code or in third-party harnesses. The same post announced Haiku 5.5, available in the Claude Platform and Claude Code, which on average costs around 75% less to run than Haiku 4.5.

For prompts under 100K tokens, Haiku 5.5 costs $0.10 per million input tokens and $0.50 per million output tokens, with cache reads at $0.01. Above 100K tokens the price rises to $0.50 and $2.50 per million, with cache reads at $0.05.

ClaudeDevs suggests using it as a subagent alongside Opus 5.5 or Sonnet 5.5 for high-volume, cost-sensitive work such as summaries, compactions and database queries. At the under-100K input rate, $100 of credit buys a billion input tokens before any output is counted.

Why can't the credits pay for Claude Code?

Anthropic's FAQ says the credits are for building your own apps and agents on the Claude Platform. For that reason they don't cover interactive Claude Code sessions or extra usage after you hit your plan's limits. Your subscription still pays for Claude Code. The credit pays for API calls made with a key from the linked organization.

Think of it as two wallets. The subscription covers you working at the keyboard. The credit works like a prepaid card for the software you build. Under the June design, usage moved to extra-usage billing at full API rates once the credit ran out, if extra usage was enabled. Otherwise, Agent SDK requests stopped until the credit refreshed.

According to Digitalapplied, heavy Agent SDK users on a $20 Pro plan were using up to $300–600 of API-equivalent compute, a 15–30× subsidy, because an autonomous agent can send thousands of requests where a person sends dozens. DevToolPicks adds that Anthropic's own tools are built to maximize prompt cache hits, while third-party agents skip the cache and process their context from scratch on every call.

Two practical questions remain open in the material we have. We don't know what happens when the new balance runs out, or whether unused credit rolls over, which the June plan explicitly ruled out. In our view, the $500 cap is tight for Team plans: five Premium seats already reach it, so every seat added after that puts nothing more into the pool.

When the first deposit lands

Anthropic is rolling the credits out over a few days, and each account sees the offer only after seven days on an eligible plan, so a team that upgrades now waits at least a week. For Team owners, every deposit is recalculated from that day's seat count, so a seat added before the next cycle changes what arrives. The terms do not yet say whether unused credit carries over.